Skip to content

optimize UT for dp transfer and dp compress #610

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
amcadmus merged 3 commits into from
May 11, 2021
Merged

optimize UT for dp transfer and dp compress #610

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
6286554
Merge pull request #33 from deepmodeling/devel
denghuilu May 11, 2021
c721128
optimize UT for dp transfer
denghuilu May 11, 2021
7b5d4de
fix bug in test_transfer.py
denghuilu May 11, 2021
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
76 changes: 41 additions & 35 deletions source/tests/test_model_compression.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -11,9 +11,9 @@
if GLOBAL_NP_FLOAT_PRECISION == np.float32 :
default_places = 4
else :
default_places = 6
default_places = 10

def file_delete(file) :
def _file_delete(file) :
if os.path.exists(file):
os.remove(file)

Expand All @@ -30,9 +30,11 @@ def setUp(self):
json.dump(jdata, fp, indent=4)

ret = os.system("dp train " + self.INPUT)
assert(ret == 0), "DP train error!"
ret = os.system("dp freeze -o " + self.frozen_model)
assert(ret == 0), "DP freeze error!"
ret = os.system("dp compress " + self.INPUT + " -i " + self.frozen_model + " -o " + self.compressed_model)
assert(ret == 0), "Model compression error!"
assert(ret == 0), "DP model compression error!"

self.dp_original = DeepPot(self.frozen_model)
self.dp_compressed = DeepPot(self.compressed_model)
Expand All @@ -46,16 +48,16 @@ def setUp(self):
self.box = np.array([13., 0., 0., 0., 13., 0., 0., 0., 13.])

def tearDown(self):
file_delete(self.INPUT)
file_delete(self.frozen_model)
file_delete(self.compressed_model)
file_delete("out.json")
file_delete("compress.json")
file_delete("checkpoint")
file_delete("lcurve.out")
file_delete("model.ckpt.meta")
file_delete("model.ckpt.index")
file_delete("model.ckpt.data-00000-of-00001")
_file_delete(self.INPUT)
_file_delete(self.frozen_model)
_file_delete(self.compressed_model)
_file_delete("out.json")
_file_delete("compress.json")
_file_delete("checkpoint")
_file_delete("lcurve.out")
_file_delete("model.ckpt.meta")
_file_delete("model.ckpt.index")
_file_delete("model.ckpt.data-00000-of-00001")

def test_attrs(self):
self.assertEqual(self.dp_original.get_ntypes(), 2)
Expand Down Expand Up @@ -163,9 +165,11 @@ def setUp(self):
json.dump(jdata, fp, indent=4)

ret = os.system("dp train " + self.INPUT)
assert(ret == 0), "DP train error!"
ret = os.system("dp freeze -o " + self.frozen_model)
assert(ret == 0), "DP freeze error!"
ret = os.system("dp compress " + self.INPUT + " -i " + self.frozen_model + " -o " + self.compressed_model)
assert(ret == 0), "Model compression error!"
assert(ret == 0), "DP model compression error!"

self.dp_original = DeepPot(self.frozen_model)
self.dp_compressed = DeepPot(self.compressed_model)
Expand All @@ -179,16 +183,16 @@ def setUp(self):
self.box = None

def tearDown(self):
file_delete(self.INPUT)
file_delete(self.frozen_model)
file_delete(self.compressed_model)
file_delete("out.json")
file_delete("compress.json")
file_delete("checkpoint")
file_delete("lcurve.out")
file_delete("model.ckpt.meta")
file_delete("model.ckpt.index")
file_delete("model.ckpt.data-00000-of-00001")
_file_delete(self.INPUT)
_file_delete(self.frozen_model)
_file_delete(self.compressed_model)
_file_delete("out.json")
_file_delete("compress.json")
_file_delete("checkpoint")
_file_delete("lcurve.out")
_file_delete("model.ckpt.meta")
_file_delete("model.ckpt.index")
_file_delete("model.ckpt.data-00000-of-00001")

def test_1frame(self):
ee0, ff0, vv0 = self.dp_original.eval(self.coords, self.box, self.atype, atomic = False)
Expand Down Expand Up @@ -282,9 +286,11 @@ def setUp(self):
json.dump(jdata, fp, indent=4)

ret = os.system("dp train " + self.INPUT)
assert(ret == 0), "DP train error!"
ret = os.system("dp freeze -o " + self.frozen_model)
assert(ret == 0), "DP freeze error!"
ret = os.system("dp compress " + self.INPUT + " -i " + self.frozen_model + " -o " + self.compressed_model)
assert(ret == 0), "Model compression error!"
assert(ret == 0), "DP model compression error!"

self.dp_original = DeepPot(self.frozen_model)
self.dp_compressed = DeepPot(self.compressed_model)
Expand All @@ -298,16 +304,16 @@ def setUp(self):
self.box = np.array([19., 0., 0., 0., 13., 0., 0., 0., 13.])

def tearDown(self):
file_delete(self.INPUT)
file_delete(self.frozen_model)
file_delete(self.compressed_model)
file_delete("out.json")
file_delete("compress.json")
file_delete("checkpoint")
file_delete("lcurve.out")
file_delete("model.ckpt.meta")
file_delete("model.ckpt.index")
file_delete("model.ckpt.data-00000-of-00001")
_file_delete(self.INPUT)
_file_delete(self.frozen_model)
_file_delete(self.compressed_model)
_file_delete("out.json")
_file_delete("compress.json")
_file_delete("checkpoint")
_file_delete("lcurve.out")
_file_delete("model.ckpt.meta")
_file_delete("model.ckpt.index")
_file_delete("model.ckpt.data-00000-of-00001")

def test_1frame(self):
ee0, ff0, vv0 = self.dp_original.eval(self.coords, self.box, self.atype, atomic = False)
Expand Down
104 changes: 77 additions & 27 deletions source/tests/test_transfer.py
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -1,10 +1,10 @@
import os,sys,platform,shutil,dpdata
import os,sys,platform,shutil,dpdata,json
import numpy as np
import unittest

from deepmd.env import tf
from deepmd.infer import DeepPot
from common import tests_path
from common import j_loader, tests_path
from infer.convert2pb import convert_pbtxt_to_pb
from deepmd.entrypoints.transfer import load_graph, transform_graph

Expand All @@ -14,10 +14,52 @@
else :
default_places = 10

def _file_delete(file) :
if os.path.exists(file):
os.remove(file)

class TestTransform(unittest.TestCase) :
def setUp(self):
convert_pbtxt_to_pb(str(tests_path / os.path.join("infer","deeppot.pbtxt")), "deeppot.pb")
convert_pbtxt_to_pb(str(tests_path / os.path.join("infer","deeppot-1.pbtxt")), "deeppot-1.pb")
self.data_file = str(tests_path / os.path.join("model_compression", "data"))
self.original_model = str(tests_path / "dp-original.pb")
self.frozen_model = str(tests_path / "dp-frozen.pb")
self.transferred_model = str(tests_path / "dp-transferred.pb")
self.INPUT = str(tests_path / "input.json")
jdata = j_loader(str(tests_path / os.path.join("model_compression", "input.json")))
jdata["training"]["training_data"]["systems"] = self.data_file
jdata["training"]["validation_data"]["systems"] = self.data_file
jdata["model"]["descriptor"]["seed"] = 1
jdata["model"]["fitting_net"]["seed"] = 1
with open(self.INPUT, "w") as fp:
json.dump(jdata, fp, indent=4)

# generate the original input model
ret = os.system("dp train " + self.INPUT)
assert(ret == 0), "DP train error!"
ret = os.system("dp freeze -o " + self.original_model)
assert(ret == 0), "DP freeze error!"

# generate the frozen raw model
jdata = j_loader(str(tests_path / os.path.join("model_compression", "input.json")))
jdata["training"]["training_data"]["systems"] = self.data_file
jdata["training"]["validation_data"]["systems"] = self.data_file
jdata["model"]["descriptor"]["seed"] = 2
jdata["model"]["fitting_net"]["seed"] = 2
_file_delete(self.INPUT)
with open(self.INPUT, "w") as fp:
json.dump(jdata, fp, indent=4)
ret = os.system("dp train " + self.INPUT)
assert(ret == 0), "DP train error!"
ret = os.system("dp freeze -o " + self.frozen_model)
assert(ret == 0), "DP freeze error!"

# generate the transferred output model
ret = os.system("dp transfer -O " + self.original_model + " -r " + self.frozen_model + " -o " + self.transferred_model)
assert(ret == 0), "DP transfer error!"

self.dp_original = DeepPot(self.original_model)
self.dp_transferred = DeepPot(self.transferred_model)

self.coords = np.array([12.83, 2.56, 2.18,
12.09, 2.87, 2.74,
00.25, 3.32, 1.68,
Expand All @@ -31,37 +73,45 @@ def setUp(self):
self.expected_v = np.array([-2.912234126853306959e-01,-3.800610846612756388e-02,2.776624987489437202e-01,-5.053761003913598976e-02,-3.152373041953385746e-01,1.060894290092162379e-01,2.826389131596073745e-01,1.039129970665329250e-01,-2.584378792325942586e-01,-3.121722367954994914e-01,8.483275876786681990e-02,2.524662342344257682e-01,4.142176771106586414e-02,-3.820285230785245428e-02,-2.727311173065460545e-02,2.668859789777112135e-01,-6.448243569420382404e-02,-2.121731470426218846e-01,-8.624335220278558922e-02,-1.809695356746038597e-01,1.529875294531883312e-01,-1.283658185172031341e-01,-1.992682279795223999e-01,1.409924999632362341e-01,1.398322735274434292e-01,1.804318474574856390e-01,-1.470309318999652726e-01,-2.593983661598450730e-01,-4.236536279233147489e-02,3.386387920184946720e-02,-4.174017537818433543e-02,-1.003500282164128260e-01,1.525690815194478966e-01,3.398976109910181037e-02,1.522253908435125536e-01,-2.349125581341701963e-01,9.515545977581392825e-04,-1.643218849228543846e-02,1.993234765412972564e-02,6.027265332209678569e-04,-9.563256398907417355e-02,1.510815124001868293e-01,-7.738094816888557714e-03,1.502832772532304295e-01,-2.380965783745832010e-01,-2.309456719810296654e-01,-6.666961081213038098e-02,7.955566551234216632e-02,-8.099093777937517447e-02,-3.386641099800401927e-02,4.447884755740908608e-02,1.008593228579038742e-01,4.556718179228393811e-02,-6.078081273849572641e-02])

def tearDown(self):
os.remove("deeppot.pb")
os.remove("deeppot-1.pb")
_file_delete(self.INPUT)
_file_delete(self.original_model)
_file_delete(self.frozen_model)
_file_delete(self.transferred_model)
_file_delete("out.json")
_file_delete("compress.json")
_file_delete("checkpoint")
_file_delete("lcurve.out")
_file_delete("model.ckpt.meta")
_file_delete("model.ckpt.index")
_file_delete("model.ckpt.data-00000-of-00001")

def test(self):
self.graph0 = load_graph("deeppot.pb")
self.graph1 = load_graph("deeppot-1.pb")
new_graph_def = transform_graph(self.graph1, self.graph0)
with tf.gfile.GFile("deeppot-2.pb", mode='wb') as f:
f.write(new_graph_def.SerializeToString())
self.dp = DeepPot("deeppot-2.pb")
os.remove("deeppot-2.pb")
ee, ff, vv, ae, av = self.dp.eval(self.coords, self.box, self.atype, atomic = True)
ee0, ff0, vv0, ae0, av0 = self.dp_original.eval(self.coords, self.box, self.atype, atomic = True)
ee1, ff1, vv1, ae1, av1 = self.dp_transferred.eval(self.coords, self.box, self.atype, atomic = True)
# check shape of the returns
nframes = 1
natoms = len(self.atype)
self.assertEqual(ee.shape, (nframes,1))
self.assertEqual(ff.shape, (nframes,natoms,3))
self.assertEqual(vv.shape, (nframes,9))
self.assertEqual(ae.shape, (nframes,natoms,1))
self.assertEqual(av.shape, (nframes,natoms,9))
self.assertEqual(ee0.shape, (nframes,1))
self.assertEqual(ff0.shape, (nframes,natoms,3))
self.assertEqual(vv0.shape, (nframes,9))
self.assertEqual(ae0.shape, (nframes,natoms,1))
self.assertEqual(av0.shape, (nframes,natoms,9))
self.assertEqual(ee1.shape, (nframes,1))
self.assertEqual(ff1.shape, (nframes,natoms,3))
self.assertEqual(vv1.shape, (nframes,9))
self.assertEqual(ae1.shape, (nframes,natoms,1))
self.assertEqual(av1.shape, (nframes,natoms,9))
# check values
for ii in range(ff.size):
self.assertAlmostEqual(ff.reshape([-1])[ii], self.expected_f.reshape([-1])[ii], places = default_places)
for ii in range(ae.size):
self.assertAlmostEqual(ae.reshape([-1])[ii], self.expected_e.reshape([-1])[ii], places = default_places)
for ii in range(av.size):
self.assertAlmostEqual(av.reshape([-1])[ii], self.expected_v.reshape([-1])[ii], places = default_places)
for ii in range(ff0.size):
self.assertAlmostEqual(ff0.reshape([-1])[ii], ff1.reshape([-1])[ii], places = default_places)
for ii in range(ae0.size):
self.assertAlmostEqual(ae0.reshape([-1])[ii], ae1.reshape([-1])[ii], places = default_places)
for ii in range(av0.size):
self.assertAlmostEqual(av0.reshape([-1])[ii], av1.reshape([-1])[ii], places = default_places)
expected_se = np.sum(self.expected_e.reshape([nframes, -1]), axis = 1)
for ii in range(nframes):
self.assertAlmostEqual(ee.reshape([-1])[ii], expected_se.reshape([-1])[ii], places = default_places)
self.assertAlmostEqual(ee0.reshape([-1])[ii], ee1.reshape([-1])[ii], places = default_places)
expected_sv = np.sum(self.expected_v.reshape([nframes, -1, 9]), axis = 1)
for ii in range(nframes, 9):
self.assertAlmostEqual(vv.reshape([-1])[ii], expected_sv.reshape([-1])[ii], places = default_places)
self.assertAlmostEqual(vv0.reshape([-1])[ii], vv1.reshape([-1])[ii], places = default_places)